Are ferrite cable shields on coax a good idea?

[Nemo]

I've got some very low level signals at up to 150kHz going down some
RG174 cable, which is pretty well shielded (85% coverage or more if I
recall the cable spec correctly). It struck me I might be able to reduce
noise further by adding a ferrite tube to the cable, which I've often
done with unshielded wires. However, when asked to explain this to a
colleague, I found I didn't know as much as I thought I did, and
couldn't convey how ferrites act as common mode chokes to him.

On thinking about how coax works, with the inner and outer current
cancelling each others' fields, I wondered if adding a ferrite might
actually *disrupt* the flow of current round the coax and make
interference worse rather than better - i.e. by increasing impedance to
high frequency currents so that the coax shield was poorer at high
frequency. And reading further about ferrite cable shields, they seem to
be discussed mainly in terms of reducing emissions rather than reducing
incoming interference. Another thing that makes me wonder if mixing
ferrites and coax is bad is: articles discussing coax as a solution to
interference make no reference to ferrite cores. I am now rather
confused on the issue. If anyone has any experience in combining
ferrites and coax I'd appreciate advice on this matter.

 

[Baron]

Perhaps the source of confusion is the notion that the choke will
open-circuit the shield at AC, leaving the coax vulnerable to pickup
along its whole length. That isn't so, because for even mode signals
it open-circuits the centre conductor as well, and for odd mode
signals it doesn't do anything.

Cheers,

Phil Hobbs

Hi Phil,
I was with you up to the last paragraph !
Since the currents on the inside of a co-axial cable are separate from
any current on the outside. How can a ferrite tube affect current on
the inner conductor ?

 

[Joerg]

I've got some very low level signals at up to 150kHz going down some
RG174 cable, which is pretty well shielded (85% coverage or more if I
recall the cable spec correctly). It struck me I might be able to reduce
noise further by adding a ferrite tube to the cable, which I've often
done with unshielded wires. However, when asked to explain this to a
colleague, I found I didn't know as much as I thought I did, and
couldn't convey how ferrites act as common mode chokes to him.

On thinking about how coax works, with the inner and outer current
cancelling each others' fields, I wondered if adding a ferrite might
actually *disrupt* the flow of current round the coax and make
interference worse rather than better - i.e. by increasing impedance to
high frequency currents so that the coax shield was poorer at high
frequency. And reading further about ferrite cable shields, they seem to
be discussed mainly in terms of reducing emissions rather than reducing
incoming interference. Another thing that makes me wonder if mixing
ferrites and coax is bad is: articles discussing coax as a solution to
interference make no reference to ferrite cores. I am now rather
confused on the issue. If anyone has any experience in combining
ferrites and coax I'd appreciate advice on this matter.

A ferrite dampens the current on anything that goes through it in the
same fashion. So if you run the whole coax through there it will not
disturb the differential signal (which you want to preserve) but will
muffle asymmetrical currents (which you seem to want to get rid of). You
said up to 150khz but didn't say how low it goes. Looks like a 77 or J
material toroid would be great. The thicker the better, but also the
more expensive. You can slide several of those over the cable to improve
things some more. If you must muffle stuff in the kHz range as well
you'd also need an old transformer that you can scrape out and run the
coax through.

Oh, and make sure to secure the ferrite. Ferrite is brittle, if it
clangs against something hard it can shatter.

 

[Jeroen Belleman]

I've got some very low level signals at up to 150kHz going down some
RG174 cable, which is pretty well shielded (85% coverage or more if I
recall the cable spec correctly). It struck me I might be able to reduce
noise further by adding a ferrite tube to the cable, which I've often
done with unshielded wires. However, when asked to explain this to a
colleague, I found I didn't know as much as I thought I did, and
couldn't convey how ferrites act as common mode chokes to him.
[...]

I could go on for hours on this subject. Here's my take on the
matter:

The desired signal over coax is purely odd-mode, i.e., the current
flowing over the inner conductor is accompanied by an equal and
opposite current flowing along the inside of the screen.

Interfering signals ('noise') are even mode. If the cable screen
had been perfect, these would not affect the desired signal. The
screen conductor however has a finite, non-zero, resistance, and
so, at low frequencies, an even mode signal will leak into the
cable.

There are two strategies to reduce this leakage. One seeks to
reduce the ingress of interference by lowering the screen
resistance: Use double-screened cable or double up the screen
by running a copper braid or thick cable in parallel with it.
The other seeks to lower even mode currents by reducing the
interfering source (if it can be identified) or by increasing
the impedance of the loop formed by the cable screen and the
-unknown- return path. (Not by inserting resistance into the
screen, of course! That would be the worst you could do!)

Slipping ferrites over the cable falls in that last category.
The trouble is that the loop referred to in the previous
paragraph usually has an inductance of the order of 10uH or
(much) more. Slipping a ferrite over your cable is not going
to increase that substantially, and therefore the effect is
likely to be minimal.

In an application of these ideas in a particle accelerator
here, I managed to gain an additional 20dB of interference
rejection on some position pick-ups in the machine. This
enabled us to observe beams with a factor of ten lower
intensity, which is just what was needed to measure the
probe beams for the LHC.

Jeroen Belleman

 

[MooseFET]

I've got some very low level signals at up to 150kHz going down some
RG174 cable, which is pretty well shielded (85% coverage or more if I
recall the cable spec correctly). It struck me I might be able to
reduce noise further by adding a ferrite tube to the cable, which I've
often done with unshielded wires. However, when asked to explain this
to a colleague, I found I didn't know as much as I thought I did, and
couldn't convey how ferrites act as common mode chokes to him.
[...]

I could go on for hours on this subject. Here's my take on the
matter:

The desired signal over coax is purely odd-mode, i.e., the current
flowing over the inner conductor is accompanied by an equal and
opposite current flowing along the inside of the screen.

Interfering signals ('noise') are even mode. If the cable screen
had been perfect, these would not affect the desired signal. The
screen conductor however has a finite, non-zero, resistance, and
so, at low frequencies, an even mode signal will leak into the
cable.

There are two strategies to reduce this leakage. One seeks to
reduce the ingress of interference by lowering the screen
resistance: Use double-screened cable or double up the screen
by running a copper braid or thick cable in parallel with it.
The other seeks to lower even mode currents by reducing the
interfering source (if it can be identified) or by increasing
the impedance of the loop formed by the cable screen and the
-unknown- return path. (Not by inserting resistance into the
screen, of course! That would be the worst you could do!)

Slipping ferrites over the cable falls in that last category.
The trouble is that the loop referred to in the previous
paragraph usually has an inductance of the order of 10uH or
(much) more. Slipping a ferrite over your cable is not going
to increase that substantially, and therefore the effect is
likely to be minimal.

In an application of these ideas in a particle accelerator
here, I managed to gain an additional 20dB of interference
rejection on some position pick-ups in the machine. This
enabled us to observe beams with a factor of ten lower
intensity, which is just what was needed to measure the
probe beams for the LHC.

Jeroen Belleman

Ferrite can help with RF, but not with ground loops.  Also, whether
pickup is purely even mode or not depends on how the two conductors are
loaded.

How about RF ground loops?

 

[Jeroen Belleman]

[...]
Interfering signals ('noise') are even mode. If the cable screen
had been perfect, these would not affect the desired signal. The
screen conductor however has a finite, non-zero, resistance, and
so, at low frequencies, an even mode signal will leak into the
cable. [...]

Jeroen Belleman

Ferrite can help with RF, but not with ground loops. Also, whether
pickup is purely even mode or not depends on how the two conductors are
loaded.

Cheers,

Phil Hobbs

I have this gadget in the lab to measure cable leakage. Basically
it's a brass pipe with a piece of cable mounted on-axis inside it,
so that the space between the cable screen and the pipe wall also
is an approximately 50 Ohm transmission line. So we have a coax
inside a coax. All four ends are looking into 50 Ohms.
The idea is to inject a signal between the cable screen and the
pipe wall, and then look how much of it gets picked up on the inner
line.

For very low frequencies (<100kHz, I may be off a bit; This was 7
years ago), this shows that the leakage path is simply a resistive
divider, with the cable screen resistance and the termination
impedance as the main players. So for a 1m piece of RG223, with
about 7 mOhm of screen resistance, I get about -77dB of ingress.
Above 100kHz, the ingress drops by 20dB/decade, until is gets into
the analyzer noise floor at about -110dB.

At the high frequency end, at some point, leakage would go up again
because of the incomplete coverage of the screen. I cannot see that
however. My setup doesn't work very well beyond 200MHz, because it's
not easy to cleanly couple a signal into the outer line.

Maybe I should dig this out again. Many people might be interested
in some cable leakage data. It's not something commonly detailed
in cable specs. (It's more fun than writing reports, which is what
I should be doing right now.)

Jeroen Belleman

 

[Jeroen Belleman]

On Jan 23, 3:17 am, Phil Hobbs


How about RF ground loops?

Aha! The infamous ground loop. Whatever you choose to call it,
the thing that matters is current flowing over the cable screen.
Normally the screen is grounded at both ends. There may be
something that forces a potential difference, not necessarily DC,
between the two ends. That will cause a current to flow in the
screen and since the screen isn't perfect, some of it will find
its way into your signal.

The nature of the remedy will depend on the coupling mechanism
and the frequency of the agressor signal. The agressor source
and the way it couples are often difficult to identify, if only
because the 'components' of the equivalent circuit are not on
the schematic diagram and of unknown magnitude. Once you get
a model of the way the interference gets into your box, the
fix is usually obvious.

Jeroen Belleman

 

[Nemo]

Thank you everyone, loads and loads of useful info there.

To answer your concerns, I don't think ground loops aren't a problem,
but I'll look into that. I usually secure ferrites to cables with some
heat shrink. I suspect the lowest frequency of interest is about 5kHz -
I'll need to check the system specs there, but certainly not DC.

 

[Joerg]

Thank you everyone, loads and loads of useful info there.

To answer your concerns, I don't think ground loops aren't a problem,
but I'll look into that. I usually secure ferrites to cables with some
heat shrink. I suspect the lowest frequency of interest is about 5kHz -
I'll need to check the system specs there, but certainly not DC.

For 5kHz you'd have to loop the cable several times through the ferrite.
Or just use an old audio transformer core or small line transformer core.

 

[MooseFET]

Aha! The infamous ground loop. Whatever you choose to call it,
the thing that matters is current flowing over the cable screen.
Normally the screen is grounded at both ends.

Unless you loop both of the cable's ends back to exactly the same
place, you can't really say that both ends are grounded without
qualifications. Consider a cable that is run between two points that
are 1/2 a wavelength apart at some frequency. The shield in the cable
is never perfect and there are always external sources of radio waves
in the environment.

If the COAX in question is on the order of 10,000 feet and is running
from down in the ocean and onto a ship, "RF" starts to mean some
fairly low frequencies.

 

[Frithiof Jensen]

Thank you everyone, loads and loads of useful info there.

To answer your concerns, I don't think ground loops aren't a problem, but
I'll look into that. I usually secure ferrites to cables with some heat
shrink. I suspect the lowest frequency of interest is about 5kHz - I'll
need to check the system specs there, but certainly not DC.

For frequencies that low you may be better off with an "iron" core or maybe
a iron powder toroid.

 

[MooseFET]

For frequencies that low you may be better off with an "iron" core or maybe
a iron powder toroid.

Remember that the core is intended to remove things that are
unwanted. The frequencies of the unwanted currents is what really
matters.

 

[Jeroen Belleman]

Normally the screen is grounded at both ends.

Unless you loop both of the cable's ends back to exactly the same
place, you can't really say that both ends are grounded without
qualifications. [...]

Agreed. That was my point. 'Ground' isn't universal. Between two
different points called 'ground' there is always some impedance
and depending on conditions, there may be a sizable voltage.

Jeroen Belleman

 

[Ross Herbert]

:On Jan 25, 2:47 am, "Frithiof Jensen"
:> "Nemo" <[email protected]> skrev i
meddelelsen:>
:> > Thank you everyone, loads and loads of useful info there.
:>
:> > To answer your concerns, I don't think ground loops aren't a problem, but
:> > I'll look into that. I usually secure ferrites to cables with some heat
:> > shrink. I suspect the lowest frequency of interest is about 5kHz - I'll
:> > need to check the system specs there, but certainly not DC.
:> > --
:> > Nemo
:>
:> For frequencies that low you may be better off with an "iron" core or maybe
:> a iron powder toroid.
:
:Remember that the core is intended to remove things that are
:unwanted. The frequencies of the unwanted currents is what really
:matters.


Yes, and the ferrite on the coax is intended to prevent the shield of the cable
radiating EMI/RFI from the equipment it is connected to where it can be picked
up by other sensitive equipment. The ferrite is not there to prevent the
equipment it is connected to from picking up stray noise.

 

[Eeyore]

Only if you need them.

Graham

 

[Eeyore]

Oh, and make sure to secure the ferrite. Ferrite is brittle, if it
clangs against something hard it can shatter.

Read 'almost certainly will shatter'.

How about iron powder cores ? Not enough permeability I suppose.

Graham

 

[MooseFET]

:> "Nemo" <[email protected]> skrev i
meddelelsen:>
:> > Thank you everyone, loads and loads of useful info there.
:>
:> > To answer your concerns, I don't think ground loops aren't a problem, but
:> > I'll look into that. I usually secure ferrites to cables with some heat
:> > shrink. I suspect the lowest frequency of interest is about 5kHz - I'll
:> > need to check the system specs there, but certainly not DC.
:> > --
:> > Nemo
:>
:> For frequencies that low you may be better off with an "iron" core or maybe
:> a iron powder toroid.
:
:Remember that the core is intended to remove things that are
:unwanted.  The frequencies of the unwanted currents is what really
:matters.

Yes, and the ferrite on the coax is intended to prevent the shield of thecable
radiating EMI/RFI from the equipment it is connected to where it can be picked
up by other sensitive equipment. The ferrite is not there to prevent the
equipment it is connected to from picking up stray noise.

I have used them to prevent the equipment they are on from picking up
noise. If you have a large signal source nearby and a coax that is
some significant part of a wavelength long, you can end up with large
RF currents flowing in the shield and coupling into the core of the
coax.

 

[Archimedes' Lever]

Read 'almost certainly will shatter'.

How about iron powder cores ? Not enough permeability I suppose.

Fucking idiot. Put a piece of heat shrink over the ferrite after it is
installed onto the cable. Duh! Fucking Donkeys are so stupid.

 

[Archimedes' Lever]

I have used them to prevent the equipment they are on from picking up
noise. If you have a large signal source nearby and a coax that is
some significant part of a wavelength long, you can end up with large
RF currents flowing in the shield and coupling into the core of the
coax.

When you get up into GHz ranges, they are referred to as 'spurs'.

One reason why there are not very many switchers used in these bands.
Linear supplies make a friggin' rack module pretty heavy.

We need a switcher that has a super-smooth output that has no HF
switcher noise included with it. So what? Super caps, computer grade
cap banks?

How do we reduce switcher noise so that they do not pose a problem with
GHz range RF bands?

 

[Ross Herbert]

:> On Sun, 25 Jan 2009 12:18:33 -0800 (PST), MooseFET <[email protected]>
wrote:
:>
:> :On Jan 25, 2:47 am, "Frithiof
:>
:> :> "Nemo" <[email protected]> skrev i
:> meddelelsen:> :>
:> :> > Thank you everyone, loads and loads of useful info there.
:> :>
:> :> > To answer your concerns, I don't think ground loops aren't a problem,
but
:> :> > I'll look into that. I usually secure ferrites to cables with some heat
:> :> > shrink. I suspect the lowest frequency of interest is about 5kHz - I'll
:> :> > need to check the system specs there, but certainly not DC.
:> :> > --
:> :> > Nemo
:> :>
:> :> For frequencies that low you may be better off with an "iron" core or
maybe
:> :> a iron powder toroid.
:> :
:> :Remember that the core is intended to remove things that are
:> :unwanted.  The frequencies of the unwanted currents is what really
:> :matters.
:>
:> Yes, and the ferrite on the coax is intended to prevent the shield of the
cable
:> radiating EMI/RFI from the equipment it is connected to where it can be
picked
:> up by other sensitive equipment. The ferrite is not there to prevent the
:> equipment it is connected to from picking up stray noise.
:
:I have used them to prevent the equipment they are on from picking up
:noise. If you have a large signal source nearby and a coax that is
:some significant part of a wavelength long, you can end up with large
:RF currents flowing in the shield and coupling into the core of the
:coax.


I dare say that a ferrite core around a coax cable might achieve some degree of
"bulk" suppression of noise pick-up but this would not be optimal. The preferred
method of suppressing pickup in sensitive equipment is to do so on each of the
individual input lines at the input. This can be achieved using filtered input
connectors (if the extra expense warrants it) or simply passing the input lines
through individual ferrite beads as close to the input point as possible. Good
quality filtered connectors may include feedthrough capacitors and an inductor
in pi configuration on each pin.